Instead of using a thread for every goroutine, Go multiplexes goroutines across multiple threads ("M:N scheduling"). So instead of each thread having a default 2MB stack, each goroutine has a tiny 2KB stack that's managed by the runtime instead of the operating system. When the program needs to grow the stack for a goroutine and there's not enough room, the runtime copies the entire goroutine's stack to another place in memory where it has enough room to expand.
Challenges of BPF Tracing Go